See the same service from every network that matters.
Deploy NATS-only workers in private locations, assign monitors, set a failure quorum, and use the inter-location mesh to expose broken paths between sites.
Self-hosted / Kubernetes-first / AGPL-3.0
Monitoring without blind spots.
Watch endpoints, protocols, databases, hosts, and scripted browser journeys from public or private networks—then carry confirmed failures through configurable alerts, incidents, and status-page updates.
- Type
- Self-hosted blackbox monitor
- Protocols
- 17 monitor types
- Transport
- NATS JetStream
- State
- PostgreSQL
- Deploy
- Docker Compose · Helm
- License
- AGPL-3.0 open source
Continuous record · 4 monitors · 4 locationsRecording
14:32 · quorum 2/4 → alert → incident #482 → status page14:32 · incident #482
CH 01checkout-api
CH 02prod-postgres
CH 03login-journey
CH 04eu-edge
17monitor types across web, network, data, hosts, and journeys
Quorumlocation-aware state that distinguishes down from degraded
Livestatus-page updates over NATS and server-sent events
Own ityour infrastructure, telemetry, credentials, and retention policy
One operating picture
From first signal to final update.
Probara connects the parts that usually live in separate tools: check execution, distributed location health, alert routing, incident response, dependency context, and public communication.
Alert on state, not noise.
Tune consecutive-failure and latency-anomaly sensitivity, route per monitor or by workspace defaults, and mute named maintenance or quick snoozes. Deliver through email, Slack, Discord, Teams, or signed HTTPS webhooks—with delays, reminders, and optional incidents.
Status pages that stay in the loop.
Group services into sections, publish incident updates, theme the built-in page, or own the entire Go template with draft preview, versioned publish and revert, and a reusable template library.
Context for the failure, not just the symptom.
Model service dependencies, annotate downstream alerts, explore co-firing signals, and ask the configured LLM for incident root causes.
Built for shared operations.
Separate tenants, control writes with roles and scoped API keys, connect OIDC, retain an audit trail, and configure encryption for stored monitor and channel secrets.
17 monitor types
Check the protocol, not a proxy for it.
Verify availability at the layer users and systems actually consume—from HTTP body, header, JSON, latency, and TLS-expiry assertions to extracted multi-step API variables, scripted browser journeys, database authentication, AMQP, ICMP, DNS records, host telemetry, and dead-man heartbeats.
Web & API
HTTPWebSocketSynthetic APISynthetic Browser
Network
PingDNSgRPCTCPSIP
Databases & brokers
PostgreSQLMySQLRedisMongoDBRabbitMQ
Infrastructure & organization
Host agentPush heartbeatGroup roll-up
System design01 02 03 04
Scale the work, keep one source of truth.
Stateless Go services divide responsibilities cleanly. PostgreSQL holds the application model; NATS JetStream carries durable work and events; workers grow independently with check volume.
Inputs
Web UI & API clients
CRUD, run-now, imports, configuration, and response workflows.
Host agents & pushes
Machine telemetry and externally initiated heartbeat results.
Private workers
Location-scoped execution with NATS-only result transport.
Probara control planeHorizontally scalable
API · Scheduler · Workers
Create state, schedule due checks, execute work, and persist results.
PostgreSQL + NATS
Durable application state, JetStream queues for check jobs and alerts, and core NATS for live updates.
Alerter · Status page
State-driven response and a separately served public experience.
Outcomes
Dashboard & analytics
Live state, trends, grouped services, history, and detailed results.
Alerts & incidents
Deduplicated lifecycle, routing, timelines, and analysis.
Public status
Cached pages, incidents, maintenance, and live SSE refresh.
Schedule
The scheduler selects due monitors and publishes location-aware check jobs through NATS JetStream.
Execute
Horizontally scalable workers run protocol, database, browser, and private-network checks, while agents and push endpoints report passive results.
Evaluate
Persisted results drive monitor state, quorum, alert lifecycle, dependency context, and incidents.
Communicate
Route notifications, publish incident updates, and refresh public status pages live.
Deployment
Your stack. Your boundaries.
Start the complete development stack with one command, run application services locally against containerized infrastructure, or install the included Helm chart in Kubernetes.
Host application data and stored credentials inside infrastructure you control.
Bring external PostgreSQL and NATS or deploy the bundled dependencies.
Scale worker replicas separately and enable the Helm worker HPA.
Add OIDC SSO, scoped API keys, secret encryption, and private locations.
probara / quick start
# Clone and prepare configuration git clone https://github.com/yassinebenameur/probara.git cd probara cp .env.example .env # Set ADMIN_JWT_SECRET (32+ random characters) and PUBLIC_BASE_URL in .env # Local Go services + Docker infrastructure make start-all-local # Or Docker-backed application services make start-all # Verify the main Go module make test make lint
Own the signal. Shorten the incident.
Deploy the complete stack, connect the systems that matter, and keep the operational picture in your hands.